Scribbles

Links I'm into: papers, tools, videos, and rabbit holes, with a one-line note on why. Updated whenever something sticks.

The Bitter Lesson by Rich Sutton

Re-read this every few months. Compute plus general methods beat clever hand-crafted ones. Keeps me humble about my own 'clever' pipelines.

Where every architecture diagram and half-baked idea starts. The hand-drawn look somehow makes it okay for things to be wrong.

The benchmark behind a lot of my xulab work. Notes on why bio reasoning is a genuinely different beast from generic QA.

Interactive essays that teach systems by letting you poke them. The north star for how I wish more ML papers were presented.